Product Rating Prediction Using Centrality Measures In Social Networks
Keywords
cluster coefficients; hubs; power-law; recommendations; Social networks
Abstract
Online recommendation systems provide useful information to users on various products and also allow the users to rate the products. However, they do not usually consider the fact that users trust their connections more than others and that the trusts vary from connection to connection i.e., we value the opinions of our connections differently. Moreover, the importance of connections' opinion changes over time. Thus, there is a need to consider the evolving trust relationships among users. In this work, we use both the user's social connections and non-connections to predict how a user would rate a particular product. We argue that we not only trust our connections more but also the trust varies over time, which we capture using a time-dependent trust matrix. We use the degree and eigen-vector centrality measures in conjunction with the user-item rating matrix to find how the social connections impact how one rates a product. To test the validity of the proposed framework, we use Epinions dataset which provides the ratings for products and trust matrix over 11 time periods. We show the accuracy our predictive model using the mean absolute error.
Publication Date
11-10-2015
Publication Title
2015 36th IEEE Sarnoff Symposium
Number of Pages
94-98
Document Type
Article; Proceedings Paper
Personal Identifier
scopus
DOI Link
https://doi.org/10.1109/SARNOF.2015.7324650
Copyright Status
Unknown
Socpus ID
84966582972 (Scopus)
Source API URL
https://api.elsevier.com/content/abstract/scopus_id/84966582972
STARS Citation
Davoudi, Anahita and Chatterjee, Mainak, "Product Rating Prediction Using Centrality Measures In Social Networks" (2015). Scopus Export 2015-2019. 2082.
https://stars.library.ucf.edu/scopus2015/2082